ML Tech Lead for the Delivery Robot Planning Team

Delivery robots are a step into the future of logistics and autonomous technology. We are creating a system where thousands of robots will be able to move independently around the city, interact with people and infrastructure, delivering orders quickly and safely.

You will have a rare chance to participate in creating a fundamentally new robot module at an early stage. We are looking for a Tech Lead who will take full responsibility for one of the subdomains, build the process, and bring solutions to the fleet. We really love and support the short cycle of "idea → concept → test site → production", we have a lot of data and enough GPUs for experiments. You will have to solve a wide range of tasks: from collecting a dataset and training a planning model to optimizing the solution for working on real robots with limited resources. Our team is responsible for the robot's movement brain: from scene understanding and dynamics prediction to trajectory generation.

On Habr, you can read an overview article about the design of our robot.

What tasks await you

Lead role and managing projects involving multiple teams You will bring complex things to a reasonable logical conclusion, build work with related teams, raise the bar of engineering culture, and assist in hiring.

Implementing SOTA approaches for AV planning You will adapt promising approaches from articles for our target production model based on transformers. You will need to work with data: it's important to learn how to select complex scenes with narrow passages, a large number of agents near the robot, etc.

Model interpretability We expect ideas from you on developing model introspection tools for debugging complex cases and, ultimately, obtaining tools for prompting behavior. At the same time, it is important that the tools scale to a fleet of a thousand robots.

More about ML at Yandex — in the channel Yandex for ML

We expect that you

• Have worked as a Tech Lead in production ML systems
• Have a strong foundation in DL
• Have worked with PyTorch
• Have integrated ML solutions into real products

Will be a plus

• Have worked with ROS
• Can write industrial-grade C++ code